{"type":"link","version":"1.0","title":"Rewarding the reflection process rather than the correct answer trains more generalizable self-improvement","author_name":"AI Archs","author_url":"https://ai-arch.pages.dev","provider_name":"AI Archs","provider_url":"https://ai-arch.pages.dev","url":"https://ai-arch.pages.dev/n/rewarding-reflection-tokens-not-answers-generalizes-self-improvement","thumbnail_url":"https://ai-arch.pages.dev/android-chrome-512x512.png","thumbnail_width":512,"thumbnail_height":512}